MBO-YOLO: An Enhanced YOLOv8n-Based Algorithm for Detecting Marine Benthic Organisms

Imaging quality in complex underwater environments is often degraded by medium scattering effects, making it difficult to extract key visual features such as color, texture, and shape of marine benthic organisms. Additionally, these organisms typically exhibit small-object characteristics, carrying limited distinguishable information, which further complicates detection tasks. To address these challenges, this study proposes an marine benthic organism detection algorithm-MBO-YOLO. Built upon the YOLOv8n architecture, MBO-YOLO incorporates a P2 small-object detection layer to enhance the extraction and detection of fine-grained features. Furthermore, a Non-Local Block attention mechanism is introduced to improve the model's focus on critical regions, optimizing small-object detection performance. Additionally, a Bidirectional Feature Pyramid Network (BiFPN) is integrated to strengthen deep feature fusion, enhancing detection robustness. Experimental results on the URPC2020 dataset demonstrate that, compared to the baseline YOLOv8n model, MBO- YOLO achieves improvements of 1.0% in precision (P), 2.1 % in recall (R), 2.2% in mAP<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">50</inf>, and 2.9% in mAP<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">50:95</inf>, with only a slight increase in model complexity. These findings validate the effectiveness of MBO- YOLO in detecting blurred, densely distributed, and small-scale targets, highlighting its potential for widespread applications in marine benthic organism detection.